SERIES PAPER DISCUSSION
IZA DP No. 4225
Attitudes toward Uncertainty among the Poor: Evidence from Rural Ethiopia Alpaslan Akay Peter Martinsson Haileselassie Medhin Stefan T. Trautmann June 2009
Forschungsinstitut zur Zukunft der Arbeit Institute for the Study of Labor
Attitudes toward Uncertainty among the Poor: Evidence from Rural Ethiopia Alpaslan Akay IZA and University of Gothenburg
Peter Martinsson University of Gothenburg
Haileselassie Medhin University of Gothenburg
Stefan T. Trautmann TIBER and CentER, Tilburg University
Discussion Paper No. 4225 June 2009
IZA P.O. Box 7240 53072 Bonn Germany Phone: +49-228-3894-0 Fax: +49-228-3894-180 E-mail:
[email protected]
Any opinions expressed here are those of the author(s) and not those of IZA. Research published in this series may include views on policy, but the institute itself takes no institutional policy positions. The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center and a place of communication between science, politics and business. IZA is an independent nonprofit organization supported by Deutsche Post Foundation. The center is associated with the University of Bonn and offers a stimulating research environment through its international network, workshops and conferences, data service, project support, research visits and doctoral program. IZA engages in (i) original and internationally competitive research in all fields of labor economics, (ii) development of policy concepts, and (iii) dissemination of research results and concepts to the interested public. IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper should account for its provisional character. A revised version may be available directly from the author.
IZA Discussion Paper No. 4225 June 2009
ABSTRACT Attitudes toward Uncertainty among the Poor: Evidence from Rural Ethiopia* We looked at risk and ambiguity attitudes among Ethiopian peasants in one of the poorest regions of the world and compared their attitudes to a standard Western university student sample elicited by the same decision task. Strong risk aversion and ambiguity aversion were found with the Ethiopian peasants. Ambiguity aversion was similar for peasants and students, but peasants were more risk averse. Testing for the effect of socio-economic variables on uncertainty attitudes showed that poor health increased both risk and ambiguity aversion.
JEL Classification: Keywords:
D81, C93, O12
risk attitudes, ambiguity attitudes, poverty, cultural differences
Corresponding author: Stefan T. Trautmann Tilburg Institute for Behavioral Economics Research (TIBER) Tilburg University P.O. Box 90153 5000 LE Tilburg The Netherlands E-mail:
[email protected]
*
Financial support from Sida (Swedish International Development Cooperation Agency) to the Environmental Economics Unit at the University of Gothenburg is gratefully acknowledged.
In decision under uncertainty it has been shown that agents distinguish between prospects for which they have a clear probability assessment or feel competent because of their own expertise, and prospects for which probabilities are unknown and the agents feel less competent (Chow and Sarin, 2002; Fox and Tversky, 1995; Heath and Tversky, 1991; Viscusi and Chesson, 1999; Viscusi and Magat, 1992, Zeckhauser, 2006). The extreme case of objectively known probabilities (e.g. tails come up in a coin flip) is called risk, and the extreme case of completely unknown probabilities (e.g. rain tomorrow) is called ambiguity. Ellsberg (1961) suggested that people often prefer to bet on risky prospects instead of ambiguous prospects, even if normative theory (Savage, 1954) implies indifference. Confirming Ellsberg’s conjecture, ambiguity aversion has been found in many empirical studies, also under market conditions and with monetary incentives (e.g. Halevy, 2007; Maffioletti and Schmidt, 2005; Sarin and Weber, 1993). Most experiments on ambiguity are conducted using university students in a laboratory environment. Few studies have used non-student samples. For example, Viscusi and Chesson (1999), Maffioletti and Santoni (2005), and Cabantous (2007) show ambiguity effects for small business owners, union delegates, and insurance managers, suggesting its importance in real world decisions. A significant number of decisions under uncertainty is also made by farmers in developing regions of the world, who, in contrast to the above mentioned groups, live near or below the poverty line. The objective of this paper is to measure attitudes toward uncertainty among small scale farmers in one of the poorest regions in the world. Knowledge of attitudes toward uncertainty is important to understand choices in traditional production activities involving well known risks, as well as to understand the uptake and adaptation of new production technologies (e.g. fertilizer) and investments (e.g. water harvesting) that involve unknown risks, leading to ambiguity. While uncertainty has been identified as an important determinant of farm technology adoption (Feder, 1980; Feder et al.,1985; Kebede, 1992), the literature does not systematically differentiate between the effect of risk aversion and ambiguity aversion. Although ambiguity attitudes have been widely observed, there is little direct evidence in the context of development and subsistence farming. Henrich and McElreath (2002) conducted an experimental study of risk and ambiguity attitude among Chilean Mapuche small scale farmers and found no evidence for ambiguity aversion. They argued that ambiguity aversion may be driven by cultural factors, and that it does not generalize to non-western farming societies. Their explanation is consistent with the finding of a strong
1
social component in ambiguity attitude (Curley et al., 1986; Trautmann et al., 2008a): pursuing ambiguous options for which little information about the risk is available is perceived as poor decision making and people seem to anticipate this and therefore avoid ambiguous risks. Such a social stigma for poor information options may not exist in nonwestern culture, reducing ambiguity aversion. Henrich and McElreath’s study shows the problems of generalizing attitudes under uncertainty from standard experimental participant pools toward culturally different groups that are of economic interest. Their findings, however, may not generalize toward other farming societies either. Two points of concern with their results are that (1) the Mapuche were unusually risk seeking for known probabilities, and (2) that there was no control experiment among a standard participant pool for the ambiguity experiment. It is conceivable that the Mapuche held especially optimistic attitudes to the chance events, possible due to friendly long-term relations with the experimenters, and therefore had optimistic views concerning the distribution of payoffs in the ambiguous task. Situations under ambiguity are even more sensitive to framing than situations of known risks (Maffioletti and Schmidt 2005; Trautmann et al. 2008b). Small differences in the decision tasks compared to studies with standard student samples at Western universities could therefore provide an obvious explanation for the observed ambiguity attitudes in the absence of a student control group. This paper measures attitudes toward uncertainty among small scale farmers in rural Ethiopia. Our subjects differ from standard subject pools in terms of their occupation, wealth, and cultural background. The experiment used real monetary incentives with high stakes and compared the results to data from university students in the Netherlands facing the same decision tasks. The measure of ambiguity aversion controls for individual differences in risk attitudes, and we relate both risk and ambiguity attitude to socio-economic variables. The next section gives a description of the participant pool and introduces the experimental design. The results are presented in Section 2 and discussed in Section 3. The last section offers concluding remarks.
1. Participants and experimental design
Participants. The experiment was conducted in the village of Abraha We Atsbaha in the northern highlands of Ethiopia. The majority of the Ethiopian population resides in the highlands, where small-scale subsistence agriculture is the main economic activity. Highland agriculture in Ethiopia is characterized by population pressure, extreme land fragmentation,
2
severe soil degradation, and heavy dependence on rainfall. As a result, the overall outcome is one of the lowest agricultural productivity levels in the world. During the last few decades, the number of droughts has exacerbated the problem, especially in the northern parts of the country. Abaraha We Atsbaha is one of many very poor villages in a region where most people depend on food aid programs to survive between the two annual harvests. Our sample consisted of 92 adults with little or no formal education, and 30% of those who participated in our experiment were illiterate. Subjects were randomly selected from a list of 584 households, with either the male or female household head participating. All subjects were peasants and mainly growing wheat, maize, barley and teff. Most families also own some livestock like cattle and sheep. All participants were Christians.
Payoffs. Each participant could win up to 20 Ethiopian birr (ETB). At the time of the experiment the exchange rate was ETB 9.67 = US$ 1, and the 2007 purchasing power parity conversion rate was 6.02 for Ethiopia (Source: IMF World Economic Outlook Database, April 2008). Note that in this region, the daily wage for unskilled farm labor varies between 10–15 Birr depending on the season.
Procedure. We elicited each participant’s certainty equivalents for a risky and an ambiguous prospect using a choice list. The risky prospect allowed the participant to bet on the color of a ball drawn from a bag with exactly 5 white and 5 yellow balls to win ETB 20 if the color is guessed correctly. Thus, this prospect offers a 50% chance to win the prize. The ambiguous prospect allowed participants to bet on the color of a ball drawn from a bag with 10 balls, where each ball can be either white or yellow. The proportion of colors in the ambiguous bag is unknown. If the color is correctly predicted the participant wins ETB 20. The two prospects are the risky and ambiguous option in the Ellsberg (1961) twocolor choice task. The ambiguous option is always at least as good as the risky option. If the participant is indifferent between betting on either color in the ambiguous option, she should be indifferent between betting on the risky option and betting on the ambiguous option. She should therefore have identical certainty equivalents for both options. If the participant believes that there are more white balls than yellow balls in the ambiguous bag, she will bet on white in the ambiguous prospect and should prefer this prospect to the risky prospect because it gives her a higher chance to win the prize, given her believes. A similar argument holds if the participant believes that there are more yellow balls in the ambiguous bag.
3
To elicit participants’ certainty equivalents we offer them 20 choices between a sure payoff and playing the prospect. Choices are arranged in a choice list. The sure payoff increases from ETB 1 to ETB 20. For very small sure payoffs, most participants will prefer to play the prospect, for very large sure payoffs most participants will prefer the sure cash. That is, most participants will switch from sure cash to playing the prospect at some point. We calculate the certainty equivalent as the midpoint between the lowest sure payoff for which the participant takes the sure cash and the highest sure payoff for which the participant prefers to play the prospect (see illustration in the appendix). Note that this choice list methodology differs from the list employed by Binswanger (1980), where participants were asked to choose one prospect from a list of prospects that differed with respect to their expected payoff and variance, and the selected prospect then served as an index of risk aversion. Our method is closer to Henrich and McElreath’s (2002) by directly eliciting the certainty equivalent of one prospect, but it avoids the chained procedure that these authors use. In the chained procedure the decision problems that participants are offered depend on their previous choices. With the choice list all participants face the same decision problems. Also, in contrast to chained procedures, the choice list is incentive compatible. Participants make choices in one choice list for each prospect. That is, in total they made 40 choices. After the participants made all choices, one of these choices was randomly selected for real play for each participant. Depending on his decision in the selected choice problem, the participant received either the sure cash amount or played the prospect with a chance to win ETB 20. Because most of our subjects had no formal education and many were illiterate, the instructions were given verbally in local language using posters as visual aids. All probabilities and randomizations were demonstrated using balls and dice, and no explicit reference to probabilities was given. Visual aids have been shown to improve the understanding of risks by participants without formal training in probability theory and were clearly necessary in our sample (Carlsson et al., 2004; Corso et al., 2001). The prospects and the betting tasks were demonstrated using the risky option by filling the bag with 5 white and 5 yellow balls. A subject chose a color by putting a ball of this color on the table. Next, a ball was randomly drawn. If the colors matched, the subject was paid ETB 20. The binary choices between the prospects and the sure amounts of money were presented to the participant by the experimenter one at a time. The experimenter filled out the choice list according to the participant’s preference in each choice.
4
Control group. As a comparison standard, we use data from an experiment where the above decision task was conducted with undergraduate university students at a Dutch university (Trautmann et al., 2008b, experiment 4). The tasks and randomizations were identical to the Ethiopian sample apart from the following. The prize was €50 for the two prospects for the student sample, and 2 of 79 students were randomly selected for real play of their choices. Students received written instructions and filled out the choice lists by themselves.
2. Experimental results
2.1 Risk and ambiguity attitudes
Risk attitudes. The certainty equivalents for the risky prospect allow us to control for risk attitude in the measurement of ambiguity below. Risk attitudes are of independent interest, however, and we report the data here. In this subsection we assume expected utility with power utility and report constant relative risk aversion (CRRA) coefficients. With the simple two-outcome gain prospects studied here the results do not change if we assume linear utility and interpret risk aversion in terms of probability weighting as in rank dependent utility and prospect theory5. The median coefficient of relative risk aversion in the Ethiopian sample is =0.73, which is significantly larger than the median of =0.34 in the Dutch student sample (MannWhitney U test, z=4.391, p
